2084 I n d . E n g . Chem. Res.

1995,34,2084-2093

SEPARATIONS

Modeling of a Packed Column Contactor

Abu Bakr S. H. Salem*
Chemical Engineering Department, King Fahd University of Petroleum & Minerals,
Dhahran 31261.Saudi Arabia

Mohammad S.H. Al-Fandery and Abdul Al Haq M. Abdul Al Latif

Department of Chemical Engineering & Technology, College of Technological Studies, Kuwait

A 5.5 cm diameter and 125 cm long glass column, packed with 10 mm glass Raschig rings was
used as a liquid-liquid contactor. The contactor performance was investigated by the system
acetic acid-toluene-water a t 25 "C.The data obtained have been compared with previous works
using mixer-settler contactors with the same test system. A new modeling approach is used to
simulate extraction in packed columns. This approach is based on dividing the packing section
into a number of strips simulating the stages in tray contactors. A nonequilibrium stage model
is developed in which equilibration of the phases is not necessarily attained. Material balance
relationships have to be only strictly observed. The model results are in excellent agreement
with the experimental data. An extraction efficiency parameter similar to the vaporization
efficiency developed by 'Holland is introduced. This parameter is given by the ratio of the actual
composition of the extract to that of the raffinate phases obtained from the contactor divided by
the equilibrium ratios. The extraction efficiency is in effect the fractional equilibrium achieved
in a n actual situation. Using the nonequilibrium stage model may result in better simulation
of the actual performance of the commercial contactors and in the design of this equipment.

Introduction ingly, significant differences in efficiency can arise,

Various types of liquid-liquid extraction contactors
have been developed during the past few decades. The
user is faced with a critical choice between contactors.
Gourdon et al. (1991) proposed an objective criterion t o
be used for comparison of two types of mechanically
agitated extraction columns. These criteria are based
on the local effects (transport, breakup) that the drops
undergo and the effects determining the residence time,
the interfacial area, and, therefore, the global efficiency
of the contactor involved.
A logical criterion for comparing different types of
contactors may be obtained by using these contactors
in performing a certain extraction duty. In such a case,
the test system, the residence time, and the output
raffinate stream specified for each contactor should be
the same. The performance of each contactor can be
evaluated by using a suitable model. Molstad et al.
(1942,1943) adopted a similar approach in investigating
the performance of drip-point grid packing compared
with other types of packing.
However, in this connection, many older liquid-liquid
extractors are being reevaluated to take advantage of
newer, more efficient contacting devices. This has
created a need for more reliable flooding and mass
transfer correlations for commercial scale columns, since
previous correlations have been verified only with data
from small diameter columns. Large commercial con-
tactors are characterized by a low surface area to
volume ratio. Under these conditions, the type of drop
movement may deviate substantially from that occur-
ring in smaller diameter laboratory columns. Accord-
* Author to whom correspondence should be addressed.
0888-5885/95/2634-2084$09.0QlO 0 1995 American Chemical Society
which might not be predicted by presently available
mathematical models.
The present work is aimed at developing a math-
ematical model to simulate the behavior of packed
column extractors. The packing bed is divided into a
number of elements of equal height. This number is
assumed to be equivalent to the number of stages
usually included in a tray column. Equilibrium condi-
tions are not necessarily attained. An extraction ef-
ficiency parameter is proposed to account for the
behavior of liquids used, the design of the stage, or the
packing section and the prevailing operating conditions.
In other words, the simulation technique is based on
adopting a nonequilibrium stage model where partial
equilibrium is only achieved. This is usually expected
in commercial contactors.
An experimental study using a packed column is
presented. The model results obtained were in excellent
agreement with the experimental findings. A compari-
son between the packed column contactor and mixer-
settlers is discussed.
Experimental Equipment
A flow diagram for the extraction unit UOP 5 used
in this investigation is shown in Figure 1. The equip-
ment is mounted on a floor-standing welded steel frame
fitted with adjustable feet. This unit includes an
extraction column, a distillation unit for the solvent
recovery, and a control panel. The extraction column
consists of a Pyrex glass tube of 55 mm internal
diameter, and 1250 mm height. The column is fitted
with a top and bottom enlarged sections of glass bulbs
of about 380 mm diameter. These bulbs serve as

12
4
I -
8
2
4 I
1:Extraction Column 2,3:Feed tank
6:CentrifugaI Pump 7 : Solvent tank
13: Solenoid Valve 1 4 , l S : S w i t c h intheControl Panel.
Figure 1. Flow diagram of the extraction system.
receivers for the input streams and also for disengaging
the raffinate and extract phases. The truncated sections
of the bulbs are closed by stainless steel plates. The
lower plate is bolted to the framework to support the
column. The column proper and the two end plates are
all fastened together with flanges. The joints between
the sections of the column are sealed with molded PTFE
gaskets.
The column was filled with packing rings supported
on a perforated stainless steel plate which is fitted
between the bottom enlarged portion and the lower
section of the column. This plate contains 20 holes of 7
mm diameter. A layer of 0.37L of 10 mm Raschig rings
was placed on the top of a smaller layer of 0.1 L of 15
mm rings. The two layers form the packing bed in the
column. Three stainless steel storage tanks for the
liquids are mounted on the same framework. The light
phase was allowed to flow from its supply tank by using
a centrifugal pump through an air bleed valve, a flow
control valve, and a flowmeter to an injector. The
injector was mounted on the baseplate. The exit of the
injector delivers the liquid at about 150 mm above the
plate. This phase leaves the top of the column through
a pipe and is collected in a polythene receiver tank. The
heavy phase supply tank provides the feed through a
diaphragm meter pump. The pumping rate of this type
is varied by a stroke adjustment knob. A digital readout
shows this rate as a percentage of the maximum flow.
The pumped liquid enters the top of the column via an
injector similar to that fitted at the base of the column
for the light liquid. A cock for sampling and draining
is fitted in the liquid feed line. The heavy liquid is
withdrawn from the bottom of the column and returned
t o a receiver vessel via a pipeline which is fitted with a
solenoid valve and a sampling cock.
The level of the light-heavy phase interface in the
column is determined by the operation of a solenoid
valve in the heavy phase outlet pipeline. The operation
of this valve is controlled by two sets of water-sensing
Ind. Eng. Chem. Res., Vol. 34,No. 6, 1995 2085
-8-
$1
I
I
4: E x t r a c t Receiver 5 : Metering Pump
8: Raffinate Receiver 9,10, I I , 12 : Flawmeters
electrodes; one is fitted to the top plate and the other is
fitted to the bottom one. The system control panel is
provided with a switch which selects the electrodes in
use and hence determines whether the interface is at
the top or the bottom of the column. Each set consists
of three stainless steel electrodes. One bare earth
electrode makes continuous contact with the light liquid.
The two other electrodes are insulated up to 5 mm apart
from their ends. These insulated electrodes are set to
heights differing by 5 mm t o produce a liquid level
differential of 5 mm. The purpose of this arrangement
is to avoid frequent opening and closing of the solenoid
valve which would occur with a simple single-electrode
system. The electrical sensing system operates at a low
ac voltage which provides a small value of the conduc-
tion current through the light liquid. The latter being
translated into a solenoid-operating voltage by a plug-
in module behind the control panel.
The distillation column boiler mounted behind the
extraction column is fitted at such a height that liquid
may be drained into it from the highest of the three
tanks and can be drained from it into the lowest one. A
valve controls the release of the solvent into the boiler.
The boiler, pipework, valves, and fittings are con-
structed in stainless steel. Heating is provided by
means of two 500 W cartridge elements inserted at the
base of the boiler. The boiler temperature is indicated
on a thermometer. The boiler lid is perforated for fitting
the distillation column. The distillation column proper
is made up of a glass section containing four sieve
plates. The glass reflux divider is bolted to the column
top and to a stainless steel condenser through flanges,
and all the sections are sealed together with molded
PTFE gaskets.
Experimental Procedure and Results
The toluene-water system was used for the non-
mass-transfer flooding and holdup studies.
2086 Ind. Eng. Chem. Res., Vol. 34, No. 6, 1995

Table 1. Solubility and Equilibrium Data of the System Water-Acetic Acid-Toluene at 25 OC

(a) Solubility Data
toluene-rich phase water-rich phase
toluene, wt % acetic acid, wt % water, wt % RI water, wt % acetic acid, wt % toluene, wt % RI
100 0 0 1.4900 100 0 0 1.33
89.30 10.63 0.07 1.4837 86.651 13.304 0.045 1.3436
85.52 14.39 0.09 1.4787 80.502 19.440 0.058 1.3474
80.13 19.72 0.15 1.4728 74.314 26.618 0.068 1.3512
77.12 22.70 0.18 1.4694 68.252 31.672 0.076 1.3551
70.32 29.25 0.43 1.4612 60.552 39.352 0.096 1.3602
65.32 33.90 0.60 1.4561 55.511 44.387 0.102 1.3634
58.49 40.71 0.80 1.4483 49.139 50.011 0.850 1.3671
52.18 44.98 2.84 1.4428 38.940 59.205 1.855 1.3728
47.97 47.84 4.19 1.4401 33.870 63.210 2.920 1.3752
42.67" 51.85" 5.48" 1.4356 22.520 70.160 7.32 1.3794
(b) Equilibrium Data
water phase toluene phase
RI Y, % RI x, %
1.3472 19.50 1.4941 1.58
1.3493 22.54 1.4936 1.78
1.3516 26.36 1.4927 2.52
1.3557 32.36 1.4918 3.60
1.3585 36.56 1.4898 4.41
1.3621 41.69 1.4886 5.52
1.3664 47.86 1.4872 7.08
1.3708 54.33 1.4848 8.81
1.3772 63.52 1.4802 12.59
1.3823 70.54 1.4792 21.59
1.3827 71.07 1.4754 24.60
a Plait point.
Table 2. Distribution Ratios of the Components of the
In the non-mass-transfer mode, the organic light System Acetic Acid (1)-Toluene (2)-Water (3)at 25 "C
phase feed was introduced and dispersed near the
bottom of the extractor in the heavy continuous water acid concn in raffinate distribution coeff
phase. The light phase was collected at the top of the phase XI, wt % Ki 1 0 4 ~ ~ K3
column to the raffinate receiver. The heavy continuous 0.500 35.770 5.056 82.615
phase was fed near the top of the column. The operating 0.700 27.130 5.209 84.281
interface was maintained at the top of the extractor by 0.980 20.797 5.417 86.905
the solenoid valve. 1.250 17.402 5.670 89.513
In the mass-transfer mode, acetic acid solute was 2.300 11.690 6.459 101.117
blended with the organic toluene feed to give a feed 2.700 10.640 6.838 106.304
3.800 8.940 7.567 108.450
mixture in the vicinity of 15.0 w t % solute. Water was 5.600 7.370 8.253 112.670
used to extract the acid from toluene using different 7.300 6.300 9.125 114.430
phase ratios and residence times. 8.500 5.930 10.234 117.560
Toluene, 99 mol % pure, QZO = 0.866 kgL, and acetic 10.200 5.350 11.341 120.780
acid, 99.75 mol % pure, QZO = 1.049 kg/L, were obtained 11.500 5.030 12.129 124.290
from the Fisher Scientific Co. Distilled deionized water 12.590 4.950 13.215 128.350
used was produced in the department. The solubility
data for the system used were determined at 25 "C by relatively fast moving or dynamic liquid holdup, hd,
the titration method which is described elsewhere which accounts for the major solute transfer; and the
(Salem, 1991). The equilibrium data were also deter- relatively stagnant or static holdup, h,, which exchanges
mined at 25 "C. Equal amounts of water and toluene mass with the surrounding dynamic liquid phase through
were stirred with a certain weight of solute for 2 h. The a convective-type diffusion process. The relative pro-
mixture was then allowed t o settle overnight in jacketed portions of h d and h, are important to the designer in
Smith-Bonner cells kept a t 25 "C. The phases were as much as they give the extent of effective liquid holdup
separated and the solute content in each phase was for mass transfer.
determined by refractometry using calibration graphs The liquid holdups in packed towers are generally
obtained from the solubility data. Compositions of the measured by either the drainage method (Shulman et
phases were determined by using an Abbe Mark 2 al., 1955) or by the tracer methods (Schubert et al.,
digital refractometer and the calibration graphs of the 1986). Kushalkar and Pangarkar (1990) reported that
respective phases. the static holdup determined by the tracer technique is
The solubility and equilibrium tie lines data are given less than that obtained by the drainage method. They
in Table 1. The distribution ratios for the system also reported that each type of the liquid holdup
components are given in Table 2. increases with the increase of the superficial liquid
velocity while the holdup ratio h&d decreases with the
Hydrodynamics increase of that velocity.
The total liquid holdup in this present work was
The liquid holdup in a packed column is comprised measured by the drainage method (Bonnet, 1982). In
of two major parts (Kan and Greenfield, 1983): the this method the valves in the feed and exit lines were
 Ind. Eng. Chem. Res., Vol. 34, No. 6, 1995 2087
Table 3. Holdup of the Solvent Phase (Water) in the Column
~ S C mass
, fraction
no packing using packing
flow rate, kglmin @SF,mass SIF, without with without with
no. t,min F S fraction kg/kg mass transfer mass transfer mass transfer mass transfer
1 2.73 0.9140 0.3400 0.27 0.375 0.270 0.300 0.29 0.304
2 2.73 0.7290 0.5480 0.43 0.750 0.310 0.330 0.33 0.350
3 2.73 0.6440 0.6440 0.50 1.000 0.350 0.351 0.36 0.375
4 2.73 0.4380 0.8760 0.67 2.000 0.370 0.353 0.39 0.400
5 5.45 0.5398 0.2025 0.27 0.375 0.680 0.700 0.71 0.774
6 5.45 0.4310 0.3240 0.43 0.750 0.700 0.710 0.72 0.806
7 5.45 0.3806 0.3806 0.50 1.000 0.710 0.740 0.76 0.830
8 5.45 0.2580 0.5170 0.67 2.000 0.740 0.760 0.78 0.860

0.8
With packing
- X Wifhout pocking

0.7-

C
.-c
-
L
ul
X 0.6-

2 0.5-

I
8 0.4-

0.3 -

0.27 I I I 1

rapidly and simultaneously closed. The phases were c ' 3 1

E
discharged, collected, and weighed. The fractional
holdup of the dispersed phase (toluene) was measured
after complete separation of the phases. The solvent
(water) holdup was obtained by difference.
The holdup measurements were conducted with and
without mass transfer using different phase ratios and
residence times. The results in case of using packing
and without packing are shown in Table 3. Figure 2
shows the solvent phase holdup at different phase ratios
and different residence times.
Flooding in packed towers occurs when true counter- 0 5 :
01
,
0 2
. ,
03
, ,
0 4
. ,
05
,
06
, , ,
07
, 1
08
current flow no longer exists. Examples of flooding
Superficial Velocllyof DiSc8rsed pnaae cmh
situations include entrainment of drops by the continu-
ous heavy phase. The dispersed light phase will be Figure 3. Flooding of the packed column contactor.
unable to penetrate packing, resulting in a buildup of
light phase below the packing support, and phase absence of packing. Figure 2 shows the variation of 4sc
inversion may occur (Seibert et al., 1990). None of these compared to SF against the solvent to feed ratio at
phenomena should be tolerated since they will eventu- different residence times. This figure shows that at z
ally upset downstream processes. = 2.73 min the change of dsc without packing assumes
In this investigation phase inversion was observed the same trend with a slight change from about 0.27 to
just when flooding was about to start. Figure 3 shows 0.37 for a change of phase ratio from 0.375 to 2.0. In
the values of the superficial velocities of the two phases the same range of phase ratios, ~ S changes
F from 0.27
at flooding conditions. to 0.67. This shows clearly that correlation between 4sc
and SF is not a straightforward one.
Discussion The figure also shows that when the residence time
was doubled, the values of 4sc were also doubled but
The holdup of the solvent phase in the column, ~ S C , assuming the same trend of change. It can be seen also
and that in the feed, ~ S F are
, shown in Table 3. The that the presence of packing has resulted in an increase
table includes the values obtained in case of mass about 5%in the holdup values of &c. The data in case
transfer and without mass transfer, in the presence and of mass transfer will show the same trend as that
2088 Ind. Eng. Chem. Res., Vol. 34, No. 6, 1995

Table 4. Results of Extraction in the Packed Column

input flow outflow
res time t, toluene, water, SIF, XO,mass x , mass y, mass x * , mass EMR, PR,
min kgimin kgimin kgkg fraction fraction fraction fraction K c/o 9%
(a) With Packing
5.45 0.457 0.1700 0.375 0.15 0.019 23.0 0.0179 12.85 99.00 87
5.45 0.653 0.2750 0.750 0.15 0.016 20.7 0.0157 13.18 99.70 89
5.45 0.322 0.3220 1.000 0.15 0.007 11.5 0.0070 16.40 100.00 95
5.45 0.218 0.4375 2.000 0.15 0.005 8.5 0.0055 15.40 100.30 98
2.73 0.914 0.3400 0.375 0.15 0.040 35.0 0.0370 9.50 97.30 73
2.73 0.729 0.5480 0.750 0.15 0.032 30.0 0.0300 10.00 98.15 78
2.73 0.644 0.6440 1.000 0.15 0.025 28.5 0.0240 11.90 98.40 83
2.73 0.438 0.8760 2.000 0.15 0.023 23.5 0.0210 11.10 99.00 86
(b) Without Packing
5.45 0.5398 0.2025 0.375 0.15 0.025 21 0.015 14.00 92.5 83
5.45 0.4310 0.324 0.750 0.15 0.021 18 0.011 16.30 92.8 86
5.45 0.3800 0.38 1.000 0.15 0.013 15 0.009 16.60 97.2 91
5.45 0.3500 0.517 2.000 0.15 0.009 11 0.007 15.70 97.9 94
2.73 1.0700 0.403 0.375 0.15 0.042 32 0.032 10.00 90.7 70
2.73 0.8480 0.648 0.750 0.15 0.040 28 0.028 10.00 90.1 73
2.73 0.7600 0.760 1.000 0.15 0.030 24 0.018 13.00 91.0 80
2.73 0.5160 1.033 2.000 0.15 0.025 22 0.016 13.79 93.3 83

observed in Figure 2. Inspection of these data shows
that the mass-transfer mechanism has resulted in about
10%increase in the holdup values of +sc.
Comparison of +sc with +SF indicates whether or not
the heavy solvent phase is accumulated in the column.
C approximately the same as that of
If the value of ~ S is
+SF, it can be assumed that the column is in continuous
smooth operation. Cases in which +sc is less than +SF
indicate that the solvent is not allowed to have a proper
residence time in the column. This can result in lower
mass transfer rates. On the other hand, when 4sc is
greater than +SF, the solvent will stay in the column
for longer periods. This may have a relatively positive
effect on the mass transfer rate. It may also delay
supplying fresh solvent to the column resulting in
solvent dead pockets, discrepancies in the hydrodynam-
ics leading to lower efficiencies of extraction.
Inspection of the data in Table 3 may highlight this
discussion point. At t = 2.73 min, and phase ratio of 1,
+SC values with or without packing are less than ~ S =
0.5. At t = 5.45 min, the values of +sc at the same phase
ratio are greater than +SF. This may indicate that
operating the column at t = 4.5 min allows no ac-
cumulation in the column and hence better separation
efficiency will be achieved.
Figure 3 shows the flooding characteristics of the
packed column contactor used. This figure shows that
at low values of dispersed organic phase superficial
velocity of about 0.15 c d s flooding of the column occurs
at a continuous water phase velocity of about 1.2 c d s .
On the other hand increasing the dispersed phase
velocity to about 0.75 c d s allows flooding to take place
at smaller rates of continuous phase of about 0.55 c d
s. However, the flooding velocities are functions of the
system physical properties, the design parameters of the
column, and the porosity of the packed section. The
values indicated by this figure determine the velocities
which should not be exceeded for safe operation of the
column without flooding. For example if the superficial
velocity of the organic phase is around 0.4 c d s , flooding
will not take place in this column if the water phase
velocity is kept usually below 0.85 c d s .
However, reference should be made here t o the works
of Molstad et al. (1942,1943)on the loading and flooding
characteristics of a packed absorber using different
types of packing. There, the loading point was defined
as that point at which the rate of change of pressure
drop with a change of gas velocity abruptly increases.
The phenomenon of flooding occurs when the gas
velocity is increased past the loading point until the
liquid is seen to begin building up above the top of the
packed space. These authors suggested that the loading
point is more significant than the flooding point as a
guide in the design of towers using the drip-point grid
packing.
Table 4 shows the results obtained when the column
was used t o extract acetic acid from toluene by using
water as a solvent at ambient temperature. Part a of
the table shows the data obtained when the column was
packed; part b shows the data obtained in absence of
packing. Two values of residence times, z, were used
in each case as shown in that table. Four different
phase ratios, SIF,starting from 0.375 to 2.0 were used
with a fured value of t. The corresponding flow rates
for the two phases are shown in the table. The feed
concentration was kept constant at 0.15 w/w. The mass
C
fraction of the raffinate and extract streams are in-
cluded. The value of the raffinate stream concentration
at equilibrium with the outlet extract stream, x * , and
the equilibrium ratios of the acid are also given in the
same table. From these values, the Murphree efficiency
of the raffinate stream can be estimated assuming that
the packed bed can be considered as a single theoretical
stage. Hence,
Table 4 shows that the efficiency of the unpacked
column at z = 2.73 min has a maximum value of 93.3%
at a phase ratio of 2. Packing increased this value to
99%. At t = 5.45 min the efficiency has increased by
packing the column from 97.9 t o about 100%. Hence it
can be assumed that the packed section height of 125
cm is equivalent to one theoretical stage. The percent-
age recovery, PR, calculated as
xo - x
PR=- x 100
XO
is also included in Table 4. The table shows that the
increase in recovery due to packing is only about 8%at
t = 5.45 min. This relatively small increase in the
recovery reveals that the presence of packing has little
 Ind. Eng. Chem. Res., Vol. 34, No. 6, 1995 2089
100
a With pocking
x without pocking

B
u
0,
+

I I I 1 I I I I
0.0 0.50 1.00 1.50 2.00
Phase Ratio W F1
Figure 4. Percentage recovery vs phase ratio ( S I F ) at different residence times.

effect on the mass transfer rate since the system used Symbol NTU HT U 'imin Pocking

has a high distribution ratio and large interfacial 1-0 1-0 273 N
x 2-0 2-b 273 V
tension. Figure 4 shows the relation between the 3-0 3-b 545 N
percentage recovery and phase ratio. The figure indi- @ 4'0 4-b 5 45 Y

cates that appreciable increase in the percentage re-

covery of about 30% can be obtained by using packing
and doubling the residence times. 1
Treybal(1973) suggested using the following formula I 700

to calculate the overall number of transfer units in a

packed column contactor based on raffinate phase
concentrations.
NTU =

where xo is the initial feed concentration and XO* is the

raffinate phase concentration at equilibrium with the
outlet extract phase. K is the slope of the equilibrium
line in the very dilute solution. R is the total flow rate
of the raffinate phase leaving the column, and S is the
total flow rate of the solvent entering the column. 05 15

Equation 3 was used to calculate the NTU using the Pnore Ratio 1S/FI

values in Table 3. The data obtained are shown in

Figure 5 plotted vs the phase ratios used. Curves l-a
and 2-a are obtained at t = 2.73 min while curves 3-a
and 4-a are those a t z = 5.45 min. The maximum value
of NTU obtained was about 3.5 at S I F = 2 for the
packed tower when t was 5.45 min.
In a perfectly mixed tray, the point efficiency would
be equal to the Murphree efficiency of the tray. In this
case, the Murphree efficiency can be estimated using
the overall number of transfer units, NTU. According
to Holland (19751, Dribika and Biddulph (1987), and
Koziol and Mackowiak (1992) the Murphree efficiency,
EMis given by
E M = 1 - exp(-NTU)
For a value of NTU = 3.5, EMR will be 0.97, which can
be considered to be in good agreement with the values
given in Table 4 with a certain error tolerance.
The height of transfer unit, HTU, was calculated by
dividing the height of the packed section by the NTU.
Figure 5. Number of transfer units and height of transfer units
vs phase ratio.
The values obtained are shown in Figure 5 on curves
l-b through 4-b. The minimum HTU obtained was
about 340 mm at a phase ratio of 2 for the tower
operating at t = 5.45 min.
It can be said that four transfer units are required
for such separation. Each unit has about 300 mm
packing height if 10 mm Raschig rings are used.
Comparison with Mixer-Settlers
Mixer-settler contactors are usually preferred in
(4) most cases t o packed towers due to their high through-
puts and higher extraction efficiency obtained. A
comparison between the performance and design pa-
rameters of the packed column contactor used in this
present investigation with two previously published
works (Salem, 1991, 1993) in which a single-stage
mixer-settler was used is given in Table 5. The same
2090 Ind. Eng. Chem. Res., Vol. 34, No. 6, 1995
Table 5. Comparison between the Packed Column Contactor and Single-Stage Mixer-Settlers Using the Same System
single-stage mixer-settler
Salem (1991) Salem (1993)
mixer two cylindrical sections cylindrical vessel
each of 5.5 cm diam, of 5.5 cm diam,
12.0 cm height 18 cm height
capacity 0.55 L 0.31 L
settler vertical tube same
capacity 0.36 L 0.395 L
total stage capacity 0.85 L 0.705 L
5 3-6 min 4.67 min
phase ratio, S I F 0.375-2.3 0.5-3.5
feed concentration 0.15 wlw 0.15 wlw
EMR 0.99 at 2000 rpm 1.00 at 1200 rpm
test system acetic acid-toluene-water was used in
these studies. The performance in the published studies
shows that the mixer-settler in both cases acted as an
ideal stage. According t o the analysis shown in Table
4,the packed column in the case of using a phase ratio
of 2 at 5.45 min residence time can be also considered
as an ideal stage. In this case, 125 cm packing height
is equivalent to a theoretical stage if 10 mm glass
Raschig ring packing is used. However, t o pursue the
analysis, another modeling approach is tried using the
concept of an actual stage.
A Mathematical Model Approach
Holland (1975, 1981) proposed that the packing bed
in a packed column used for distillation or absorption
can be divided into N sections or elements numbered
down from the top or vice versa. If equilibrium condi-
tions are attained in these sections, then N is said to
be equivalent to the number of theoretical plates, NTP,
in a column containing plates. The height of each
packing section is, therefore, equivalent to that of a
theoretical plate, HETP (Rubac, 1968).
In actual operation, stages rarely, if ever, operate at
equilibrium despite attempts to approach this condition
by proper design of the stage and choice of the operating
conditions. The usual way of dealing with departures
from equilibrium is by incorporating stage eficiency into
the equilibrium relations. Krishnamurthy and Taylor
(1985) indicated that it is with the introduction of this
stage efficiency parameter that the problems begin. The
first problem is that there are several different defini-
tions of stage efficiency: Murphree (19251, Hausen
(19531, generalized Hausen (Standart, 19651, vaporiza-
tion (Holland, 19811, and others. There is by no means
a consensus on which definition is best. The Murphree
efficiency, although the least soundly based, is the one
most widely used because it is easily combined with the
equilibrium equation.
In a c component system there are c - 1independent
component efficiencies, which for lack of anything better
have usually been taken t o be the same for all compo-
nents (King, 1980). Toor (1964) stated that the efficien-
cies of the different components are rarely equal simply
because of their varying behavior in the mass-transfer
mechanisms. This results due to coupling between the
individual concentration gradients, interaction effects,
reverse diffusion, and osmotic diffusion phenomena. A n
interesting consequence of such effects is that the
individual point efficiencies of different species are not
constrained to lie between 0 and 1. Instead they may
be found any where in the range from --m to +-m.
Horvath (1985) and Salem and Sheirah (1990) reported
experimental extraction stage efficiency to reach 1.3 in
some cases and 1.7 in others.
packed column contactor present work
column of 5.5 cm diam, 125 cm packed section,
0.47 L packing vol, 3.72 L operating capacity
with packing, 4.40 L without packing, bed vol
1.85 L, void vol 1.38 L, void fraction 0.75
flow rate 0.627-1.55 kglmin
2.73-5.45 min
0.375-2.0
0.15 wlw
1.00 with 125 cm HETP
In the present work, a nonequilibrium stage model
is developed which will rely only on material balances
and summation equations. Equilibrium relations are
assumed to be not applicable. However, other func-
tional relations between the yi and xi will be deduced
as for example
where mi,n is the functional relation between the
composition of i in the two phases. If equilibrium is
attained, mi,,,= Ki+,otherwise it is only a fraction of it.
In this sense, it is possible t o assume an extraction
efficiency, similar t o the absorption efficiency
proposed by Edmister (1949) and the vaporization
efficiency proposed by Holland (19811, which can be
calculated from
The parameter &,n embodies the effects of stage design
variables, the operating variables, and the system
variables.
In actual packed towers, the number of packing
sections, N , is that equivalent to the number of actual
stages, NAS. The height of the packing section is the
height equivalent to an actual stage, HEAS. The
algorithm which is used in this model assumes that the
number of actual stages is known.
Simulation of a Packed Column Extraction
Using a Nonequilibrium Stage Model
A sketch of a typical packed column contactor is
shown in Figure 6. The packed bed is divided into N
sections from bottom to top. The notation employed in
the description of this column corresponds to that used
in the treatment of columns with plates. The bottom
element of packing is assigned the number 1, and the
top element is given the number N . n represents any
element of packing of a height, HEAS. The feed mixture
containing the solute is introduced below the bottom
section if it is the light phase. The heavy solvent phase
is introduced from the top of the tower and flows
countercurrently to the light phase. The flow rate and
composition of these two streams are represented by Fo,
xi,o, L N + ~and
, yi~v+l,respectively. The corresponding
outlet raffinate and extract phase flow rates and con-
centrations are represented by FN, xi&, L I , and yi,1,
respectively. The raffinate phase is collected from the
top while the extract phase is discharged from the
bottom of the contactor. f i , ~f ,i ~ v ,l i ~ v + land l i , ~are the
flow rates of component i in the corresponding streams.
Usually in liquid extraction design problems the
values of Fo, xi,^, L N + ~and, yi&+l, are given and the
 Ind. Eng. Chem. Res., Vol. 34, No. 6, 1995 2091
(Treybal, 1986).

N From eq 11the values ofL, up to LN cail be estimated.

Then an overall material balance on the column gives

LN+1+ Fo = L 1 + FN (12)

This equation can be rearranged and written as

.1 LN+, - F N = L1 - Fo= A (13)
I
Fn Ln+1 H.E.A.S Another equation representing an overall material
n balance on the tower section between stage n and the
7- top stage N can be written as

Fn + L N +=~FN + L n + 1 (14)
Rearranging eq 14 using eq 13 gives the value of the
flow rate, F,, from any stage n as follows:
F, = Lnil - A (15)
The value of F N obtained from eq 15 is to be compared
with that estimated from eq 7. If the two values for F N
do not agree in a certain tolerance, the new value
obtained from eq 15 has to be used in another round of
F o , f i , o , Xi,o
f
Figure 6. Diagram of a packed extraction column.
I Lt,li,l, Yi,l
- calculation until convergence is achieved.
The flow rates li,n of the components of the extract
stream can be calculated by an equation similar to eq
11 as follows:
recoveries or the final raffinate compositions for a
certain component are specified. It is required to find
the number of stages and the raffinate and extract flow
rates and compositions of each stage.
Since liquid-liquid extraction equipment is operated
frequently in an adiabatic manner, the isothermal sum The values of yi,,, the compositions of these streams can
rates (ISR) method developed by Tsuboka and Kateaya- be calculated using
ma (1976) can be applied. This method essentially
assumes values for N , determines the exit raffinate Yi,n = li,nlLn (17)
composition, xi&, and compares it with the specified
composition until convergence. However, it can be checked that
In the present approach, N is initially assumed and
L1 can be roughly estimated assuming that the two
solvents are immiscible, hence FN can be estimated from
The flow rates fi,, and compositionsxi,, of the raffinate
FN=F- - xi'o streams can be calculated by applying a component
i = solute under consideration balance on each stage n, where
O 1 -xip
(7)
fi,n = + li,n+l - li,n (19)
Then a component balance on the tower gives
C

Ll Yi,l =L N f l Yi,N+1 + Fdci,O - FN%,N (8) (20)

Equation 8 can be also written in terms of components
flow rates as follows: The values of xi& obtained from eq 20 are to be
compared with the assumed specifications of the outlet
1.1 , l = 1.l,NNfl+ fi,o - f i & (9) raffinate stream. If these specifications are not met, a
new value of N has to be assumed. The calculation is
The total stream flow rate, L1, and its composition,y i , l repeated until this condition is satisfied within a speci-
can be estimated from fied convergence limit.
Once the value of N is fixed, through which the values
of L,, F,, yi,,, and xi,, are determined, the values of mi,,
and the extraction efficiency EL,, can be estimated.

The other stream flow rates from any stage n, L,, can (21)
be estimated using the Edmister group relationship
2092 Ind. Eng. Chem. Res., Vol. 34, No. 6, 1995

Table 6. Simulation Results for the System Acetic Acid (1)-Toluene (2)-Water (3)
n i 1 Y f X m 6 EMR
0 1 0.032700 0.1500
2 0.185191 0.8495
3 0.000109 0.0005
0.218000 1.0000
1 1 0.03177 0.06770 0.024546 0.1170 0.5786 0.1155 0.242
2 0.00013 0.00028 0.185210 0.8826 3.17 10-4 0.2641 -0.054
3 0.43737 0.93202 0.000086 0.0004 2.33 103 18.7900 -0.014
0.46927 1.00000 0.209842 1.0000
2 1 0.023616 0.05120 0.016540 0.0820 0.6244 0.1037 0.3227
2 0.000149 0.00032 0.185231 0.9177 3.49 10-4 0.3558 -0.0643
3 0.437347 0.94846 0.000064 0.0003 3.16 103 27.008 -0.013
0.461112 1.00000 0.201840 1.0000
3 1 0.015615 0.03446 0.008667 0.04460 0.7726 0.09840 0.4814
2 0.00017 0.00038 0.185251 0.95518 3.98 10-4 0.42277 -0.0858
3 0.437325 0.96516 0.000420 0.00022 4.39 103 39.90900 -0.01
0.45311 1.00000 0.193960 1.00000
4 1 0.007737 0.01740 0.00093 0.0050 3.48 0.0973 0.8977
2 0.000190 0.00040 0.18528 0.9949 4.02 x 10-4 0.7950 -0.2561
3 0.437303 0.98220 0.00019 0.0001 9.82 x 103 118.864 -0.01
0.44523 1.00000 0.18623 1.0000
5 1 0.00000 0.0000
2 0.00022 0.0005
3 0.43728 0.9995
0.43750 1.0000

It is also possible t o estimate the Murphree stage
efficiency for the raffinate phase, EMR.
The stage recovery of each component can be estimated
as follows:
This nonequilibrium stage model has been used to
simulate the extraction of acetic acid from toluene using
water solvent in the packed column used in this study.
The flow rates and compositions of the inlet streams are
Fo = 0.218 kglmin and L N +=~ 0.4375 kglmin, i = 1 =
acid, i = 2 = toluene, i = 3 = water. q o = 0.15, XZ,O =
0.8495, and X3,O = 0.0005 mass fractions. y1,”+1 = 0.0,
ypail = 0.0005, and y 3 , ~ + 1 =0.9995 mass fractions. The
raffinate output stream compositions are those obtained
actually using a packed column operating at z = 5.45
min and solvent t o feed ratio of 2 to 1. The specified
value was 3 ~ 1=, ~0.005. The equilibrium ratios for the
system components, Ki,are obtained from the experi-
mental data shown in Table 2. Initially the number of
packing sections assumed was N = 3. This gave x l , = ~ are required for this particular extraction duty.
0.0096, which did not match the experimental results.
The calculation was repeated using N = 4. Two rounds
of calculations were needed t o get to the specifications
required. The final results are given in Table 6. The
table shows the values of stream flow rates Zi,n and fi,,,
and their concentrations yi,,, and the functional
relationships mi+,the extraction efficiencies &,n, and the
Murphree efficiency for the raffinate EMRfor each
component in each section. Values for &,n greater than
1 are obtained especially for the water component while
negative values of EMRare also shown. However, for
the acid component 1, which is the solute under con-
sideration, it can be seen that EMRhas assumed the
values around 24.0, 32.0, 48.0, and 90% for the four
stages (or sections) from the feed entrance at the bottom
to the solvent entrance at the top.
This present model shows that it is possible to develop
the extraction efficiency from operational data. Once a
data bank is available for this parameter, it will be
possible to correlate it for possible use in the design of
new contactors. In fact, no attempt was made to fit the
extraction efficiency using the operating parameters due
to the complexity of some of them and also the lack of
simple quantification of others. The number of sections,
four, is also limited, which does not allow for a sound
correlation. However, reference can be made to the
works of Molstad et al. (1942, 1943) in this context.
Conclusion
A packed column was used for removing acetic acid
from toluene by water solvent. A 125 cm packed bed
using 10 mm glass Raschig rings was successful in
obtaining 98% acid recovery. Data analysis using
equilibrium stage model proved that this contactor is
almost identical t o an ideal stage. Comparison with
data obtained from single-stage mixer-settlers tested
with the same system at approximately the same
residence times shows full agreement between the
performances obtained. Analysis of the packed column
with the conventional number of transfer units indi-
cated that four of such units each of about 300 mm HTU
A new modeling approach based on the nonequilib-
rium stage model is used also for the analysis of the
data obtained in this study. Four actual stages or
packed sections each of 300 mm height are obtained
using this simulation model. The Murphree efficiencies
of these stages are low at the feed-entering sections and
become larger as the raffinate phase flows to the outlet
top section.
The new modeling approach allows analysis of actual
existing equipment without the assumption of a hypo-
thetical stage efficiency which may produce biased
information. The nonequilibrium stage model, on the
other hand, produces the actual flow rates and composi-
tion profiles through the tower based only on the

material balance considerations. The efficiencies can
then be estimated based on the actual performance.
For testing commercial scale equipment, this model
requires a set of fitted efficiencies which can be obtained
from operational data.
The extraction efficiency can be used to provide a
correlating parameter for nonequilibrium stages. How-
ever, no attempts have been made to correlate it from
the data obtained in this work since the number of
experimental points is not sufficient to produce a sound
correlation.
Nomenclature
EMR= Murphree efficiency of the raffinate phase, fractional
F = feed flow rate, kglmin
fi = flow rate of a component i in the feed phase, kglmin
FR, = fractional recovery of component i
HEAS = height equivalent to an actual stage
HETP = height equivalent to a theoretical plate
HTU = height of a transfer unit
hd = dynamic holdup
h, = static holdup
i = component
L = solvent phase flow rate, kglmin
I, = flow rate of a component i in the solvent phase, kgl
min
K, = distribution ratio of component i at equilibrium
m, = functional relation between the actual values of y,
and x,
N = number of stages or sections
n = stage or section number
NAS = number of actual stages
NTU = number of transfer units
R = raffinate phase flow rate leaving the column, kglmin
RI = refractive index
S = solvent phase flow rate entering the column, kglmin
x = feed or raffinate phase composition, mass fraction
y = solvent or extract phase composition, mass fraction
C = sum
Greek Letters
e20 = density at 20 "C,kgL
@SC = holdup of the solvent in the column, mass fraction
@SF = holdup of the solvent in the feed, mass fraction [(SI
F) + SI, k@g
A = constant defined by eq 13
t = residence time, min
E = extraction efficiency defined by eq 6
Literature Cited
Bonnet, J. C. Ph.D. Thesis, The University of Aston in Birming-
ham, UK, 1982.
Dribika, M. M.; Biddulph, M. W. Surface Tension Effects on a
Large Rectangular Tray with Small Diameter Holes, Ind. Eng.
Chem. Res. 1987,26,1489.
Edmister, W. C. Hydrocarbon Absorption and Fractionation
Process Design Methods-Part 18, Plate Efficiency. Pet. Eng.
1949,21,(2-45.
Gourdon, C.; Casamatta, G.; Angelino, H. Single Drop Experiments
with Liquid Test Systems; Away of Comparing Two Types of
Mechanically Agitated Extraction Columns. Chem. Eng. J.
1991,46,137.
Ind. Eng. Chem. Res., Vol. 34, No. 6, 1995 2093
Hausen, H. The Definition of the Degree of Exchange on Rectifylng
Plates for Binary and Ternary Mixtures. Chem. Eng. Tech.
1953,25,595.
Holland, C. D. Fundamentals and Modeling of Separation Process;
Prentice-Hall. Inc.: Endewood Cliffs. NJ. 1975.
Holland, C. D. 'Fundamentals of Multicomponent Distillation;
McGraw-Hill: New York, 1981.
Horvath, M.; Hartland, S. Mixer-Settler Extraction Column: Mass
Transfer Efficiency and Entrainment. Ind. Eng. Chem. Process
Des. Dev. 1985,24,1220.
Kan, K. M.; Greenfield, P. F. A Residence Time Model for Trickle
Flow Reactors. AIChE J . 1983,29,123.
King, C. J. Separation Processes, 2nd ed.; McGraw-Hill: New York,
1980.
Koziol, A.; Mackowiak, J. A New Method for Determining the
Efficiency of Tray Columns with Downcomers. Chem. Eng.
Technol. 1992,15,103.
Krishnamurthy, R.; Taylor, R. A Non Equilibrium Stage Model of
Multicomponent Separation Process. AIChE J . 1985,31,449,
456, 1973.
Kushalkar, K. B.; Pangarkar, V. G. Liquid Holdup and Dispersion
in Packed Columns. Chem. Eng. Sci. 1990,45,759.
Molstad, M. C.; McKinney, J. F.; Abbey, R. G. Performance of Drip-
Point Grid Tower Packings. Trans. Am. Inst. Chem. Eng. 1942,
38, 387.
Molstad, M. C.; McKinney, J. F.; Abbey, R. G. Performance of Drip-
Point Grid Tower Packings. Trans. Am. Inst. Chem. Eng. 1943,
39,605.
Murphree, E. V. Rectifying Column Calculations. Ind. Eng. Chem.
1925,17, 747.
Rubac, R. E. Determination of Vaporization Efficienciesfor Packed
Columns at Steady State Operation. Ph.D. Dissertation, Texas
A & M University, College Station, TX,1968.
Salem, A. S. Extraction in a Single-Stage Mixer-Settler. Ind. Eng.
Chem. Res. 1991,30,1582.
Salem, A.S. Analysis of a Single-Stage Mixer-Settler Performance.
Sep. Sci. Technol. 1993,28 (7) 1479.
Salem, A. S.; Sheirah, M. A. Dynamic Behavior of Mixer-Settlers.
Can. J . Chem. Eng. 1990,68 (101, 867.
Shubert, C. N.; Lindner, J. R.; Kelley, R. M. Experimental Methods
for Measuring Static Liquid Holdup in Packed Columns. AIChE
J. i986,32,i920.
Shulman, H. L.; Ultrich, C. F.; Wells, N. Performance of Packed
Columns. 1: Total, Static, and Operating Holdups. M C h E J .
1955,1, 247.
Siebert, A. F.; Reeves, B. E.; Fair, J. R. Performance of a Large
Scale Packed Liquid-Liquid Extractor. Ind. Eng. Chem. Res.
1990,29,1901.
Standart, G. L. Studies on Distillation-V. Generalized Definition
of a Theoretical Plate or Stage of Contacting Equipment. Chem.
Eng. Sci. 1985,20,611.
Toor, H. L. Prediction of Efficienciesand Mass Transfer on a Stage
with Multicomponent Systems. AIChE J . 1964,10,545.
Treybal, R. E. Liquid Extraction. Chemical Engineers' Handbook;
Perry, R. H., Chilton, C. H., Eds.; McGraw-Hill: New York,
1973; Chapter 15, pp 15-22.
Tsuboka, T.; Katayama, J. General Design Algorithm Based on
Pseudo-Equilibrium Concept for Multistage Multicomponent
Liquid-Liquid Separation Process. J . Chem. Eng. Jpn. 1976,9,
40.
Received for review November 21, 1994
Revised manuscript received February 27, 1995
Accepted March 15, 1995@
IE9406916
@ Abstract published in Advance A C S Abstracts, May 1,
1995.